From relational model to property graph: mapping data, schema, constraints and queries

From relational model to property graph: mapping data, schema, constraints and queries

연구 분야: Databases

학회: Progress in Artificial Intelligence

The graph database systems (e.g. Neo4j, TigerGraph) are becoming widely used both by researchers and practitioners in order to deal with real-life problems such as fraud detection and social network marketing. Their success is due to their graph data model which demonstrates high efficiency when representing highly connected data and performing complex analytical tasks over it. As a huge amount of relational data still exists on the Web, many works have studied the mapping of this data into graphs. The proposed models are not suitable for real-life applications for many reasons (e.g. obfuscate the original schema; rely on theoretical query languages; do not investigate the mapping of constraints). Therefore, this article proposes a complete mapping process that translates any relational database into a graph database by mapping data, schema, constraints as well as read and update queries. Our process preserves all mapping properties: information preservation, as well as semantic, query, update and monotonicity preservation. The two first properties ensure that no information nor semantics of the original data is lost during the mapping. The query (resp. update) preservation is ensured via an algorithm that translates any SQL query (resp. update) into an equivalent Cypher query (resp. update). The monotonicity ensures that updates on the relational data do not require re-computing the equivalent data graph from scratch. Finally, we conducted an experimental study to validate our results. To our knowledge, this is the first work that studies all aspects of the mapping and provides a practical solution that can be easily integrated between two systems such as MySQL and Neo4j.